The computational performance expectations set by the information technology industry will soon require an entirely new approach to interconnect architectures and implementations. In particular, many communication applications that implement large transfers of data will require larger warehouse scale datacenters with unprecedented levels of main memory capacity interconnected by high performance fabrics with an aggregate bandwidth capability in the Exabyte per second range. The physics of conventional electrical interconnect using wired connections cannot scale to meet this demand without prohibitive cost and spatial requirements.